Saturday, February 28, 2015

Trying A Couple of Strategies

Strategic Thinking

I was given a couple of interesting ideas this week about possible sub-strategies regarding MFI.  As I have a database going back month-by-month to 2006, I can sometimes back test these strategies.  The first was one I had looked at at least five years ago.  The suggestion/question was whether you would do better if you sold from your MFI portfolio stocks that had dropped a certain percentage after you bought them. The idea is to perhaps minimize your losses.

When I looked at this five years ago, the answer was "no", you were better off holding the losers and in fact, you were better off to buy more at that time.

Since 2009, I now capture in all my portfolios the 52 week low in the year I hold the stocks.  So I can look at stocks that dropped a given percentage and see what would happen if you either (a) sold at that time or (b) bought at that time.

Here is a table assuming you sold at various % drops.  Note that I have assumed you could sell at exactly 20% drop, though in reality if a stock gaps down (look at WTW yesterday) you may have to eat a larger loss:

Stock Year Do Nothing Sell at 10% Drop Sell at 20% Drop Sell at 30% Drop
2009 40.0% 18.3% 28.9% 32.9%
2010 8.3% 3.1% 7.3% 7.7%
2011 -2.0% 0.2% -1.5% -1.9%
2012 27.2% 15.2% 20.4% 23.7%
2013 24.9% 18.0% 22.7% 24.4%
2014 14.4% 16.3% 14.6% 14.7%
Total 19.5% 11.0% 15.4% 17.2%

So this shows that for every year except 2014, you'd have been better off holding a stock, whether it drops 10%, 20% or 30%.   In 2014, if a stock dropped 10%, you would be slightly better off to lock in the loss.  Now I will note that I was unable to run this table on all of my stocks.  There were some for which I did not have a 52 week low - that can happen when there is a buyout (which would bias results to be not as good as they really were) and can happen if stock is de-listed or something (did happen with some of the Chinese Reverse Merger stocks) and this might bias in other direction.

Who Will Buy?

So now what would happen if you buy MFI stocks if they have dropped 10, 20 or 30% since they were on the list?  I looked at that as well.  At first I was under-whelmed by the return. But then I realized that in my work these were not being held for a full year, but rather a fraction of a year.  I really have no way of knowing (without more work than I am willing to do) to know how long the stocks were held. All I know is the date they were sold.  My 52 week lows are just a number, I do not know the dates.  But I think it is reasonable to assume that the drops occur 1/2 way into the stock year.  In that case, the returns are for just 1/2 a year and are pretty decent.

Buy Year Buy at 10% Drop Buy at 20% Drop Buy at 30% Drop
2009 40.4% 38.7% 44.9%
2010 8.0% 2.6% 2.5%
2011 -3.3% -1.1% -0.4%
2012 23.6% 23.7% 24.0%
2013 15.6% 9.3% 4.4%
2014 -4.6% -0.5% -1.8%
Total 15.0% 12.1% 11.7%
So only in 2011 and 2014 did you actually get a negative return buying the MFI stocks after a 10/20/30% drop.  I do think this is also a function of the broader stock market. For example, 2013 was a great year for the market - so the strategy of buying a 2012 MFI stock (recall this is composed of 12 monthly portfolios from January 2012 to December of 2012) after it had dropped was probably good as the entire market pretty much rose in 2013. Also, keep in mind the 2014 number is actually just based on one monthly portfolio (January 2014) as it is the one to close.

Strategy Two

So I think we have proven that selling MFI stocks in your portfolio if they drop 10/20/30% doesn't help, my data shows it reduces your returns 2 to 6 points.  The second strategy is similar:  if a stock was on the list from a year ago and is still on the list; does it's performance from the past year predict at all the upcoming performance?

This will take a little work.  For this one, I can go all the way back to 2006.  It is simply a relatively tedious task of matching up consecutive stock years along with consecutive performances.  Hmm, let me fire up my excel worksheet (cracking fingers):

High Level.  There were 1,405 stock years in my data where a MFI stock was on both the current monthly tracking portfolio and one year ago.  My total database is 5,417 stock years. Tossing out the first year (2006) as you could not have a match and the most recent 11 portfolios (not mature) gets us down to 4,327 stock years. So roughly 1/3 of stocks were also on the list a year ago.

I split the 1,405 stock years into deciles.  A "1" are the stocks who had the worst performance the previous 12 months. A "10" would be the 140 that had the best performance the prior 12 months.

Prior Year Decile Percentage Change
1 29.7%
2 13.5%
3 5.9%
4 0.2%
5 8.3%
6 9.5%
7 15.9%
8 9.3%
9 8.2%
10 2.6%
Grand Total 10.3%

So all 1405 stocks on average were up 10.3% the second year.  But I find it very telling that stocks that did the poorest ("1" and "2") were up on average 21.6%, while stocks that were "9" and "10" were up 5.4%.  That suggests that prior 12 months performance is a material predictor.

To put the deciles into perspective, 1 and 2 are stocks down basically 30% or more the prior year.   Let us focus on these two deciles a bit further.  The answer may be biased a bit when we consider purchase year.

Purchase Year Percent Change
2007 0.4%
2008 -3.7%
2009 51.6%
2010 -3.2%
2011 10.4%
2012 22.5%
2013 86.7%
2014 91.6%
Grand Total 21.6%
So the 21.6% when looked at by year is very interesting as it has been very powerful the past two years.  But here are counts to go along with that:

Purchase Year Count
2007                                        26
2008                                        70
2009                                        36
2010                                           8
2011                                        40
2012                                        72
2013                                        27
2014                                           1
Grand Total                                      280
So there is only one 2014 data point ==> throw that out.

Split by Market Cap

We have commented many times that larger market cap stocks have fared better.  So let us split the original deciles by market cap (very simple, "large" and "small" with the goal of equal number in each bucket).

Prior Year Decile Large Small All
1 25.2% 30.9% 29.7%
2 21.4% 9.2% 13.5%
3 -0.2% 9.2% 5.9%
4 -3.9% 2.3% 0.2%
5 13.3% 1.2% 8.3%
6 11.3% 7.0% 9.5%
7 19.9% 7.3% 15.9%
8 10.2% 7.7% 9.3%
9 4.5% 14.6% 8.2%
10 10.1% -8.9% 2.6%
Grand Total 11.1% 9.5% 10.3%

So you can still see the "all" column matches the first table.  You can see if you are a small stock and you did well the prior year (a "10") the future is bleak (-8.9%).  This also suggests a better filter than just "1" and "2" stocks are "1" and "2" large stocks (roughly bigger than 700m market cap).

By Dividend Flag

Besides market cap, we know the dividend flag is a good predictor.  Let us split these deciles by dividend flag.

Hmmm.  I may not show this table!  I think I will just describe it.  If the dividend flag is 0, then the prior year decile is important. Best to go with a "1" or a "2".  If dividend flag is "1" (meaning 2.6% or greater); you are also better off going "1" and "2"; but all deciles are solid.

In Conclusion

I covered a lot of ground here.  I think we can draw the following conclusions:

  1. Don't mess with selling MFI stocks mid-term if they are doing poorly.  That has not been successful strategy.
  2. If a MFI stock has dropped; it would not be a bad strategy to buy it or consider in next tranche.  I can not measure this absolutely from my data showing 52 week lows, as I have to make simplifying assumption the opportunity to buy is at 1/2 point.  But second part of my study strongly suggests that if an MFI stock is still on the list and is down 30% from prior year, it has good prospects.
  3. Buying MFI renewing stocks that are down 30% of more in past year has led to 21.5% gain next year (2 x "normal").
  4. Adding one additional dimension to #3 (market cap >$700m) adds a couple points to the results (23.3%).
  5. Dividend Flag (>2.6% yield) AND being in "1" or "2" (recall that means down 30% or more in prior year and being on both MFI lists results in a 35% annual gain.  That being said, this is a rare occurrence.  Of my 1405 stock years, only 36 fall in this bucket.

Here are the 36 times it has happened:

Date Stock Percent Change
February-09 MDP 145%
March-09 DLX 109%
July-08 USMO 106%
December-08 HLF 106%
July-08 USMO 95%
November-08 BVF 86%
May-12 UNTD 80%
April-08 USMO 76%
October-08 BVF 72%
May-09 NTRI 69%
December-11 LPS 69%
May-09 DLX 59%
September-08 BVF 56%
October-12 DELL 54%
July-08 BVF 51%
July-08 BVF 50%
November-12 DELL 49%
December-12 DELL 39%
July-12 MANT 37%
October-08 USMO 36%
August-12 MANT 35%
October-12 MANT 34%
August-08 BVF 30%
October-11 PETS 10%
February-08 USMO 5%
January-08 BVF -3%
January-13 NSU -4%
September-08 UNTD -13%
October-12 STRA -14%
August-09 UNTD -15%
December-07 BVF -28%
July-12 STRA -41%
May-08 NOOF -55%
March-08 NOOF -71%
February-08 NOOF -72%